CODEWITHSUNDEEP
javascriptjqueryarraysperformance
jtromans
jtromans

Reputation: 4263

Most computationally efficient method to insert into an ordered array?

I have an array of arrays containing time-series data as follows:

var timeseries = [
    [Tue Dec 31 2013 00:00:00 GMT+0000 (GMT Standard Time), 3, 2],
    [Tue Dec 31 2013 01:00:00 GMT+0000 (GMT Standard Time), 1.2, 3],
    [Tue Dec 31 2013 02:00:00 GMT+0000 (GMT Standard Time), 4, 2]
]

The first item in the nested array is a javascript date.

I have a second array which contains a list of unique javascript dates, but that may or may not already exist in the timeseries array:

var newTimes = [
    Tue Dec 31 2013 00:13:00 GMT+0000 (GMT Standard Time), 
    Tue Dec 31 2013 01:00:00 GMT+0000 (GMT Standard Time), 
    Tue Dec 31 2013 01:40:00 GMT+0000 (GMT Standard Time)
]

For each item in the newTimes array, I need to check whether that date exists in the arrays stored with the timeseries array. If it does not, I would like to insert a new array in the correct place (chronological order) into the timeseries array, and - crucially - copy the other array values from the directly preceding array. For example, combining the two above array as described would yeild:

var newTimeseries = [
    [Tue Dec 31 2013 00:00:00 GMT+0000 (GMT Standard Time), 3, 2],
    [Tue Dec 31 2013 00:13:00 GMT+0000 (GMT Standard Time), 3, 2],
    [Tue Dec 31 2013 01:00:00 GMT+0000 (GMT Standard Time), 1.2, 3],
    [Tue Dec 31 2013 01:40:00 GMT+0000 (GMT Standard Time), 1.2, 3],
    [Tue Dec 31 2013 02:00:00 GMT+0000 (GMT Standard Time), 4, 2]
]

I should note both arrays are already sorted. Also, items in newTimes will exist within the range of timeseries. Finally, the original timeseries array has an entry every hour, on the hour.

I have tried a couple of different methods but I am interested in the computationally most efficient approach. I am happy to use any appropriate well adopted third party libraries such as underscore.js if helpful.

Upvotes: 2

Views: 216

Answers (1)

jtromans
jtromans

Reputation: 4263

Having considered a Binary Search (suggested by techfoobar in the comments), and the Bubble Sort (friendzis), the most optimal solution I have come up with so far is following this logic.

As opposed to using a Binary Search, consider the following.

Because we know that both arrays are already sorted; the dates within newTimes fall within the range of timeseries dates; and that the initial timeseries entries are on the hour, every hour, we can find the difference between timeseries[0][0] date and newTimes[0] in hours. This is the initial offset required to start the bubble sort approach (more like just a standard loop and slice as discussed in the comments above). This uses the Moment.js library.

Code presented below, please feel free to suggest changes:

var firstEvent = moment(newTimes[0]);
var firstTime = moment(timeseries[0][0]) ;
var offset = firstEvent.diff(firstTime, 'hours') - 1; 


for (var i = offset; i < timeseries.length;i++) { 
    while (newTimes.length > 0 && typeof moment(timeseries[i+1]) !== "undefined" && moment(newTimes[0])  > moment(timeseries[i][0]) && moment(newTimes[0])  < moment(timeseries[i+1][0])) {
        var newElement = timeseries[i].slice(0);
        newElement[0] = moment(newTimes[0]).toDate();
        timeseries.splice(i+1, 0, newElement);
        newTimes.splice(0,1);
        i++;
    } 
}

Importantly, this copes with the edge case when there are two or more entries in the newTimes array that all fall within within just one timeseries entry, for example:

var timeseries = [
    [Tue Dec 31 2013 00:00:00 GMT+0000 (GMT Standard Time), 3, 2],
    [Tue Dec 31 2013 01:00:00 GMT+0000 (GMT Standard Time), 1.2, 3]
]

var newTimes = [
    Tue Dec 31 2013 00:13:00 GMT+0000 (GMT Standard Time), 
    Tue Dec 31 2013 00:16:00 GMT+0000 (GMT Standard Time)
]

Upvotes: 1

Related Questions